8/3/2019 NERA MAY 2010
1/71
CONFIDENTIAL
May 03, 2010
An Examination of Charges forMobile Network Elements
in Israel
Report
Prepared for the Israel Ministry ofCommunications and Ministry of Finance
8/3/2019 NERA MAY 2010
2/71
CONFIDENTIAL
Project Core Team
NERA Economic Consulting
Nigel Attenborough, Director, NERA London
Christian Dippon, Vice President, NERA San Francisco
Thomas Reynolds, Economic Consultant, NERA London
Sumit Sharma, Research Officer, NERA London
Finite State Systems
Howard Cobb, Independent Consultant, Finite State Systems Ltd., London
NERA Economic Consulting
1 Front St., Suite 2600
San Francisco, California 94111
Tel: +1 415 291 1000
Fax: +1 415 291 1020
www.nera.com
NERA Economic Consulting
15 Stratford Place
London W1C 1BE
United Kingdom
Tel: +44 20 7659 8500
Fax: +44 20 7659 8501
www.nera.com
8/3/2019 NERA MAY 2010
3/71
NERA Economic Consulting i
Contents
1 INTRODUCTION AND STUDY OVERVIEW..................................................................................... 2
2 COSTING PRINCIPLES FORMOBILENETWORKELEMENTS.......................................................... 4
2.1 Definition of Long-run Incremental Cost ........................................................................ 42.2 Definition of the Relevant Increment .............................................................................. 4
2.2.1 Total-service LRIC .............................................................................................. 4
2.2.2 Pure LRIC............................................................................................................ 4
2.3 Definition of Long Run.................................................................................................... 52.4 Common Fixed Costs....................................................................................................... 62.5 General Modeling Approach............................................................................................ 7
2.5.1 Network modeling approach................................................................................ 72.5.2 Determination of network topology..................................................................... 82.5.3 Definition of forward looking.............................................................................. 9
2.6 Wholesale Access ............................................................................................................ 9
2.7 Study Process................................................................................................................. 10
3 THENERABU-LRICMODEL................................................................................................. 11
3.1 Service Demand Assessment ......................................................................................... 133.1.1 Voice and data demand equivalence.................................................................. 193.1.2 Using demand figures ........................................................................................ 21
3.2 Network Dimensioning.................................................................................................. 22
3.2.1 Model geotypes.................................................................................................. 243.2.2 Dimensioning assumptions ................................................................................ 253.2.3 Minimum coverage network .............................................................................. 26
3.2.4 Traffic network .................................................................................................. 27
3.3 Switching ....................................................................................................................... 343.3.1 Second generation networks .............................................................................. 34
3.3.2 Third generation networks ................................................................................. 35
3.3.3 Transmission...................................................................................................... 353.3.4 Other network elements ..................................................................................... 37
3.4 Network Sharing ............................................................................................................ 37
3.4.1 2G/3G sharing.................................................................................................... 37
3.4.2 Operator site sharing.......................................................................................... 373.5 Network Asset Valuation............................................................................................... 38
3.5.1 Asset unit costs .................................................................................................. 383.5.2 MEA price trends............................................................................................... 38
3.5.3 Network operating costs .................................................................................... 38
3.5.4 Indirect network assets....................................................................................... 383.5.5 Operational expenditure associated with indirect network assets...................... 39
3.5.6 License fees, royalties and frequency fees......................................................... 39
3.6 Capital Charge Determination ....................................................................................... 413.6.1 Depreciation method.......................................................................................... 41
3.6.2 Asset lives .......................................................................................................... 42
8/3/2019 NERA MAY 2010
4/71
NERA Economic Consulting ii
3.6.3 Cost of capital .................................................................................................... 433.7 Calculation of Total LRIC ............................................................................................. 43
3.8 Unit LRIC Costs ............................................................................................................ 44
4 THE MODEL RESULTS .............................................................................................................. 45
4.1 Voice Termination ......................................................................................................... 45
4.2 SMS Termination........................................................................................................... 47
4.3 MMS Termination ......................................................................................................... 494.4 Data Termination ........................................................................................................... 51
5 POLICY RECOMMENDATION..................................................................................................... 53
APPENDIX A:MODEL SENSITIVITY ANALYSES ...........................................................................A 55
APPENDIX B:DEMAND FORECAST METHODOLOGIES ................................................................. B57
APPENDIX C:COST OF CAPITAL CALCULATIONS ........................................................................ C60
8/3/2019 NERA MAY 2010
5/71
NERA Economic Consulting iii
Figures
FIGURE 1.1NERABU-LRICMODEL SERVICES INCLUDED ............................................................ 2
FIGURE 3.1OVERVIEW OFNERAS BU-LRICMODEL ................................................................. 12
FIGURE 3.2MARKET SHARES OF OPERATORS IN ISRAEL 19992014(E)........................................ 14
FIGURE 3.3PREDICTED SHARE OF 3G IN ISRAEL 20092014......................................................... 16
FIGURE 3.4SMSTRAFFIC CONVERSION PROCESS ......................................................................... 18
FIGURE 3.5MMSTRAFFIC CONVERSION PROCESS........................................................................ 19
FIGURE 3.6WDCMABUSY HOURDATA CAPACITY..................................................................... 20
FIGURE 3.7WDCMABUSY HOURVOICE AND DATA EQUIVALENCE............................................ 21
FIGURE 3.8ILLUSTRATIVE SIMPLIFIEDNETWORKTOPOLOGY GSM.............................................. 23
FIGURE 3.9TRAFFIC DIMENSIONING CALCULATION...................................................................... 29
FIGURE 3.10EXAMPLE OF GSMFREQUENCY REUSE PATTERN ..................................................... 31
FIGURE 3.112GTRANSMISSION AND SWITCHINGNETWORKTOPOLOGY ...................................... 34
FIGURE 3.123GTRANSMISSION AND SWITCHINGNETWORKTOPOLOGY ...................................... 35
FIGURE 3.13ROUTING FACTORUSENERABU-LRICMODEL ..................................................... 44
FIGURE 4.1NERABU-LRICMODEL RESULTS PERMINUTE VOICE TERMINATION 2G................ 45
FIGURE 4.2NERABU-LRICMODEL RESULTS PERMINUTE VOICE TERMINATION 3G................ 46
FIGURE 4.3NERABU-LRICMODEL RESULTS PERSMSTERMINATION 2G................................ 47
FIGURE 4.4NERABU-LRICMODEL RESULTS PERSMSTERMINATION 3G................................ 48
FIGURE 4.5NERABU-LRICMODEL RESULTS PERMMSTERMINATION 2G .............................. 49
FIGURE 4.6NERABU-LRICMODEL RESULTS PERMMSTERMINATION 3G .............................. 50
FIGURE 4.7NERABU-LRICMODEL RESULTS PERMBDATA TERMINATION 2G ....................... 51
8/3/2019 NERA MAY 2010
6/71
NERA Economic Consulting iv
FIGURE 4.8NERABU-LRICMODEL RESULTS PERMBDATA TERMINATION 3G ....................... 52
8/3/2019 NERA MAY 2010
7/71
NERA Economic Consulting v
Tables
SUMMARY RECOMMENDATION ISRAELI TERMINATION RATES (ILSAGOROTS)2010-2014 ............ 1
TABLE 2.1MODELED OPERATORS AND TECHNOLOGIES .................................................................. 9
TABLE 3.1INPUTS USED TO FORECAST SUBSCRIBERS IN ISRAEL 20092014 ................................ 13
TABLE 3.2VOICE MARKET SHARES OF OPERATORS IN ISRAEL 2009-2014(E) .............................. 15
TABLE 3.3SMSMARKET SHARES OF OPERATORS IN ISRAEL 2009-2014(E)................................. 15
TABLE 3.4MMS AND DATA MARKET SHARES OF OPERATORS IN ISRAEL 2009-2014(E).............. 15
TABLE 3.5FORECAST 2G/3GSPLIT IN ISRAEL 20092014............................................................ 16
TABLE 3.6TRAFFIC PARAMETERS ................................................................................................. 22
TABLE 3.7GEOTYPE CHARACTERISTICS ........................................................................................ 24
TABLE 3.8GEOTYPE DATA USED IN MODEL ................................................................................. 25
TABLE 3.9MINIMUM COVERAGENETWORKEXAMPLE OF CELLCOM GSM1800.......................... 27
TABLE 3.10CELL RADII FORTRAFFIC BASE STATIONS (KM)......................................................... 27
TABLE 3.11TRAFFIC CARRIED USING MICRO- AND PICOCELLS .................................................... 28
TABLE 3.12SPECIAL COVERAGE .................................................................................................... 28
TABLE 3.13EXCERPT OF ERLANG LOOKUP TABLE ........................................................................ 30
TABLE 3.14AVERAGENONWORKING INVESTMENT TIME.............................................................. 32
TABLE 3.15DESIGN UTILIZATION PARAMETERS ........................................................................... 33
TABLE 3.16LEASED LINE CHARGES IN 2009(ILS PER ANNUM) .................................................... 36
TABLE 3.17PELEPHONE MODELED AVERAGE TRANSMISSION LENGTHS (2009, KM) .................... 37
TABLE 3.18INDIRECTNETWORKASSETS % OF TOTAL DIRECTNETWORKINVESTMENT .............. 39
TABLE 3.19INDIRECT ASSET OPERATIONAL EXPENDITURE % OF INDIRECT ASSET INVESTMENT . 39
8/3/2019 NERA MAY 2010
8/71
NERA Economic Consulting vi
TABLE 3.203GLICENSE FEES ....................................................................................................... 40
TABLE 3.21ROYALTY RATES (%) ................................................................................................. 40
TABLE 3.22SPECTRUM FREQUENCY FEES (ILS)............................................................................ 41
TABLE 3.23ASSET LIVES............................................................................................................... 43
TABLE 3.24PRETAXNOMINAL COST OF CAPITAL ISRAELI MNO2009......................................... 43
TABLE 4.1NERABU-LRICMODEL RESULTS PER-MINUTE VOICE TERMINATION 20092014 ... 46
TABLE 4.2NERABU-LRICMODEL RESULTS PERSMSTERMINATION 20092014.................... 48
TABLE 4.3NERABU-LRICMODEL RESULTS PERMMSTERMINATION 20092014 .................. 50
TABLE 4.4NERABU-LRICMODEL RESULTS PERMBDATA TERMINATION 20092014 ........... 52
TABLE 5.1VOICE (PER MINUTE)POLICY RECOMMENDATION ........................................................ 53
TABLE 5.2SMS(PER MESSAGE)POLICY RECOMMENDATION........................................................ 54
TABLE 5.3POLICY RECOMMENDATION SUMMARY........................................................................ 54
8/3/2019 NERA MAY 2010
9/71
.!
NERA Economic Consulting 1
Executive Summary
NERA Economic Consulting (NERA) was commissioned by the Israel Ministry ofCommunications (MOC) and Ministry of Finance (MOF) to construct a bottom-up cost model
to examine the charges for network elements in the mobile telephony market in Israel and, if
appropriate, recommend changes to the regulated current interconnection rates.
In line with international best practices and a previous cost model constructed for the MOC andthe MOF, we constructed a bottom-up LRIC model. In deriving the input values for this model,
we solicited specific data from the Israeli operators, ranging from current demand information
to modern equivalent asset prices, operating expenses, and price trends. We also requestedmarket level data from the MOC, such as spectrum costs or the type and number of base
transceiver stations (BTSs). Although we received some data from the operators, for many
other necessary inputs to the cost model, we had to rely on international benchmark data,forecasts, or commercially and publicly available data sources. We made a conscious effort to
use all the data received from the Israeli operators unless we found the data to be inconsistent
with international benchmarks.
The modeling approach is documented in this report. Furthermore, we prepared a technical
report that further explains the working of the NERA BU-LRIC model. The model itself isprogrammed in Microsoft Excel and has been submitted to the MOC and the MOF. It is
accompanied with company-specific input sheets for Cellcom, Partner and Pelephone.
As we detail in this report, the principal finding of our study is that current voice and SMS
termination rates are significantly above current cost levels. In the summary table below, wepresent the voice and SMS termination costs from 20092014. We recommend that the MOC
set voice and SMS at average blended cost, as shown in the table below.1 Specifically, we
recommend that all termination rates be symmetric in that the same rate would apply for allnetworks regardless of whether the network is 2G or 3G or if it terminates on Cellcoms,
Partners, or Pelephones networks.
Summary RecommendationIsraeli Termination Rates (ILS Agorots)
2010-2014
2010 2011 2012 2013 2014
Voice (per minute) 4.14 3.54 3.11 2.80 2.57
SMS (per message) 0.19 0.17 0.16 0.14 0.13
1 Average blended rates represent the weighted average 2G and 3G costs for each operator, using relativetraffic proportions as weights. We then average these blended rates across all three operators to arrive at our
recommended termination rates.
8/3/2019 NERA MAY 2010
10/71
.!
NERA Economic Consulting 2
1 Introduction and Study Overview
NERA Economic Consulting (NERA) was commissioned by the Israel Ministry ofCommunications (MOC) and Ministry of Finance (MOF) to construct a bottom-up cost model
to examine the charges for network elements in the mobile telephony market in Israel and, if
appropriate, recommend changes to the regulated current interconnection rates. NERA has builta bottom-up long-run incremental cost (BU-LRIC) model, which is based on data provided by
the Israeli mobile operators where such data were provided. However, in some instances the
limited nature of the data received from the operators meant that we had to rely on other data
sources, standard engineering assumptions, and on NERAs own expertise in mobile networkeconomics and engineering. The services subject to the analysis included second generation
(2G) and third generation (3G) calls, short message service (SMS), multimedia message service
(MMS),and data termination. Specifically, the resulting NERA BU-LRIC model is capable of
calculating the LRIC of the services shown in Figure 1.1.
Figure 1.1NERA BU-LRIC Model
Services Included
Charges fo r Mobi le Netw ork Elements (Cost in g and Pr ic ing)
2G
CDMA
3G
GSM UMTS
Voice
Termination
SMS
Termination
2G
CDMA
3G
GSM UMTS
MMS
Termination
2G
CDMA
3G
GSM UMTS
Data
Termination
2G
CDMA
3G
GSM UMTS
Charges fo r Mobi le Netw ork Elements (Cost in g and Pr ic ing)
2G
CDMA
3G
GSM UMTS
Voice
Termination
2G
CDMA
3G
GSM UMTS
Voice
Termination
SMS
Termination
2G
CDMA
3G
GSM UMTS
SMS
Termination
2G
CDMA
3G
GSM UMTS
MMS
Termination
2G
CDMA
3G
GSM UMTS
MMS
Termination
2G
CDMA
3G
GSM UMTS
Data
Termination
2G
CDMA
3G
GSM UMTS
Data
Termination
2G
CDMA
3G
GSM UMTS
Additionally, with relatively minor modifications, the model is also capable of estimating thecost of wholesale (MVNO) access and national roaming.
Furthermore, it is our understanding that Israeli mobile operators were invited to submitalternative cost models using a top-down approach. Depending on whether the MNOs
submitted alternative cost estimates, the MOC requested that NERA review these models andconsolidate the results with those of the NERA BU-LRIC model. At the time of this report,
none of the Israeli operators had submitted its own model or indicated its intention to do so.Hence, this draft report only covers NERAs BU-LRIC model. It describes our modeling
approaches, assumptions, results, and sensitivity analyses.
This report addresses the following issues:
Section 2 explains the costing principles for mobile network elements, in particular, mobiletermination and access charges.
8/3/2019 NERA MAY 2010
11/71
.!
NERA Economic Consulting 3
Section 3 provides details of the NERA BU-LRIC model, including its methodology, inputdata, algorithms and equations, and scenario and sensitivity analyses.
In section 4, we present the company-specific model results. Section 5 offers our policy recommendations, including LRIC estimates, the markup over
LRIC to recover common fixed costs, and finally the mobile termination raterecommendations.
This report is accompanied by a detailed technical report and user manual.
8/3/2019 NERA MAY 2010
12/71
.!
NERA Economic Consulting 4
2 Costing Principles for Mobile Network Elements
The objective of this study is to determine the costs incurred in operating a mobile network toprovide mobile services. In line with international best practices and Israels previous cost
model, we elected to construct a BU-LRIC cost model.2 Below, we explain its conceptual
interpretation and provide details of the methodology that has been adopted.
2.1 Definition of Long-run Incremental Cost
Long-run incremental cost is defined as follows:
The concept ofincremental costis a generic cost concept, defined as the increase in afirms total costs because of some increase in output or the costs avoided if output falls.
The specification oflong run indicates that the time horizon is sufficiently long for all typesof costs to be avoidable.
LRIC is aforward-looking concept and therefore starts from the presumption that efficient,least-cost equipment and technology are used.
We provide additional clarifications of why we believe our interpretation is the most relevant in
the following subsections.
2.2 Definition of the Relevant Increment
LRIC includes all variable (i.e., volume sensitive) costs and the fixed costs specifically relevant
to the increment of output under consideration. Two options for the increment can be chosen inthe model, namely total-service LRIC and pure LRIC.
2.2.1 Total-service LRIC
The traditional definition of the increment in LRIC models is the total-serviceincrement (i.e., TSLRIC). This definition of the increment in a LRIC model ensures that
there is a consistent basis for the measurement of the costs of outgoing and incoming
calls, and it minimizes the extent of common fixed costs and the associated issues abouthow these should be allocated across services.
2.2.2 Pure LRIC
A recent European Commission (EC) Recommendation advocates an alternative pure LRIC
approach.3 In contrast to the TSLRIC approach, which includes all calls (incoming andoutgoing) in the definition of the increment, this definition of the increment includes only
incoming (i.e., wholesale terminating) calls.
2 BU-LRIC models have been used in for setting mobile termination rates in Austria, Belgium, Denmark,
France, Greece, Hungary, Norway, Sweden, UK, and by the MOC in 2003, among others.3 See the Recommendation of the 7th of May 2009 at
http://ec.europa.eu/information_society/policy/ecomm/library/recomm_guidelines/index_en.htm.
8/3/2019 NERA MAY 2010
13/71
.!
NERA Economic Consulting 5
The choice of increment (i.e., TSLRIC or pure LRIC) has a direct impact on the allocation ofcommon fixed costs. Specifically, if the TSLRIC approach is selected, the resulting costs
include all costs that are fixed and common to incoming and outgoing calls. As a result, the
average incremental cost of call termination will lie above the marginal cost of call termination,and the remaining joint and common fixed costs will be relatively small. In this instance, the
common fixed costs include the costs of the minimum coverage network, radio spectrum, and
network management systems.
In contrast, if the ECs pure LRIC approach is selected, then only the fixed costs that are
specific to terminating calls (which are relatively small) are included in the incremental cost ofcall termination. This means that the incremental cost will approach marginal cost and hence
will be lower than in the first scenario.
The implications of the pure LRIC approach for the recovery of fixed and common costs have
sparked a vigorous debate.4 For example, respondents (including national regulatoryauthorities) have questioned the justification provided for the proposed change to a pure LRIC
approach. In particular, although the EC argues for pure LRIC because it would facilitate
efficient cost recovery, it is not clear from its proposals where common fixed costs would berecovered.
Moreover, if all service prices were set on a pure LRIC basis, common costs would not be
recovered, and the company concerned would operate at a loss. Hence, if companies were
forced to operate at a loss on mobile termination, they would attempt to increase prices for
retail services. This, so-called waterbed effect, stands to defeat the EUs aim of efficient costrecovery and enhanced consumer welfare. Based on these considerations, and the fact that, to
our knowledge, no national regulatory authority has yet implemented MTRs based on a pureLRIC approach, we recommend that the increment include all (incoming and outgoing) callsconsistent with the TSLRIC approach. Notwithstanding this, the NERA BU-LRIC model has a
user-adjustable option to calculate LRICs based on either the TSLRIC or the pure LRICapproach. We discuss the sensitivity of the LRIC estimates to the choice of increment in
Appendix A to this report.
2.3 Definition of Long Run
The specification oflong run requires that the time horizon is sufficiently long for all types of
costs to be avoidable. Alternatively, long run has been defined as:
the time horizon within which the operator can undertake capital investment
or divestment to increase or decrease the capacity of its existing productiveassets. Thus a very long time horizon is observed in which all costs, including
investment capital and all costs related to network capacity, are potentially
variable with no fixed element.5
4 See the responses to the European Commissions consultation process, available from
http://ec.europa.eu/information_society/policy/ecomm/library/public_consult/termination_rates/index_en.htm.5 European Union Independent Regulators Group, Principles of implementation and best practice
regarding FL-LRIC cost modeling, as decided by the Independent Regulators Group, November 24, 2000, p. 6.
8/3/2019 NERA MAY 2010
14/71
.!
NERA Economic Consulting 6
The definition of long run, which is embedded in the NERA BU-LRIC model, assumes that allcosts are avoidable.
2.4 Common Fixed Costs
Common fixed costs are those costs that are not incremental to any one product or service.
Consequently, they are not included in LRIC. However, as discussed above, mobile operatorsmust recover common fixed costs to remain viable. Hence, consistent with international best
practices, we recommend that LRIC be marked up to enable efficiently incurred common fixed
costs to be recovered. Notwithstanding this, the NERA BU-LRIC model allows users also to
run LRIC estimates without a markup for common costs.
Two main methods exist by which common fixed costs can be recoveredequi-proportionalmark-up (EPMU) and Ramsey pricing. Under EPMU, common fixed costs are recovered in
proportion to the incremental costs of the services. Ramsey pricing, on the other hand, recoverscommon fixed costs in inverse proportion to the price elasticities of demand of differentservices.
The U.S. Federal Communications Commission (FCC) has recognized the need to take into
account those costs shared by groups of network elements and those common to all services
and elements (e.g., corporate overheads). It noted that:
Because forward-looking common costs are consistent with our forward-
looking, economic cost paradigm, a reasonable measure of such costs shall beincluded in the prices for interconnection and access to network elements.6
The FCC accepted EPMU as an appropriate basis for recovering common fixed costs butexplicitly ruled out Ramsey pricing because of concerns that it might unreasonably limit the
extent of entry into local exchange markets by allocating more costs to, and thus raising the
prices of, the most critical bottleneck inputs, the demand for which tends to be relativelyinelastic.7
The Independent Regulators Group (IRG), although recognizing that it was standard practice tomark up incremental costs so as to recover a reasonable share of common fixed costs, did not
commit itself to, or rule out, any of the possible allocation methods. More specifically, it noted
the following:
There are various methods of recovering common costs across a range ofservices. From an economic point of view distortion is minimized by recovery
of common costs according to Ramsey Pricing. This recovers common costs
from the products based on the products relative marginal cost of production
and price elasticities. However, this method of recovering common costsrequires robust and detailed information on elasticities, which is often hard to
find. The alternative is to recover common costs according to an accounting
rule. For example, if the common input were used to produce two separate,
6 Federal Communications Commission, FCC 96-325, August 1996, 694.7 Ibid., 696.
8/3/2019 NERA MAY 2010
15/71
.!
NERA Economic Consulting 7
regulated services, one simple rule would be to split the common cost equallybetween the two services. Another example would be to recover common costs
in proportion to the incremental cost of the two services. This method of
allocating costs is known as equal proportionate mark-up (EPMU).8
Given the complexity of Ramsey pricing and the difficulty of obtaining price elasticities, thismethod has not found widespread use. Based on these considerations, the NERA BU-LRIC
model employs EPMU.
We note that estimated wholesale costs exclude retail costs as these are not relevant to
providing call conveyance for interconnection. Retail costs include sales, advertising,marketing, subscriber acquisition costs, and the costs of retail subscriber billing and customer
care. In some jurisdictions, a further markup is added to allow for the recovery, at least for a
specified period of time, of stranded and legacy costs.
2.5 General Modeling Approach
The approach we have adopted in this analysis is commonly described as a bottom-up model.
Bottom-up modeling involves the construction of an engineering model to estimate the cost ofbuilding a mobile operators network, assuming that all costs are avoidable (in the long run)
and that the network is built using forward-looking technologies. This general approach
requires that additional, more specific modeling decisions have to be made. We discuss each ofthese below.
2.5.1 Network modeling approach
In common with other telecommunications cost modeling studies, we use network componentsas the building blocks of the cost model. Specifically, the NERA BU-LRIC model derives the
costs of the services based on the different components of the network, such as mobile
switching centers (MSC), base transceiver stations (BTS), and base station controllers(BSC).This approach is adopted for two principal reasons. First, a component-based approach is the
most practical approach as component costs can be identified more readily in a bottom-up
model than other building increments (such as services or network layers). Second, and moreimportantly, the costs imposed on the network by different forms of usage (e.g., mobile call
termination) are directly related to the components utilized by each of the services. If, for
example, an operator provides 2G mobile call termination interconnection to a competitor, it
must provide capacity in the tandem switch, the home location register (HLR), the MSC, theBSC, the BTS and the associated communications linkages. However, the competitor imposes
no costs on the terminating operators general packet radio service (GPRS) data network
elements.
To derive service costs based on component costs, the NERA BU-LRIC model uses so-calledrouting factors. Routing factors specify the average number of units of each network
component used by a particular type of service. Routing factors are commonly measured by the
8European Union Independent Regulators Group, Principles of Implementation and Best Practice RegardingFL-LRIC Cost Modelling, November 24, 2000, p. 5
8/3/2019 NERA MAY 2010
16/71
.!
NERA Economic Consulting 8
operators from traffic samples and are often already used as a means of establishing the cost ofretail call services. In the case of interconnection services, some of the routing factors can be
established almost by definition. For example, mobile call termination will make use of a one
BTS, while an on-net call will make use of two BTSs.
In order to model the different technologies, as shown in Figure 1.1 above, the model uses awide routing factor matrix, which includes routing factors for all the modeled technologies.
The values in this matrix are based on the operators submissions, but where necessary (e.g., if
factors were not provided or were anomalous) they have been amended and supplemented by
NERA based on telecommunications engineering knowledge and principles.
2.5.2 Determination of network topology
Another issue to be considered is whether the modeled network topology mirrors the networks
currently in place or whether the model assumes that forward-looking LRIC costs are based ona fully efficient operator that builds an ideal topology to match future demand levels. These two
topology concepts are commonly referred to as scorched node and scorched earth.Specifically, scorched node is an approach that takes the current location and number of
network nodes as the basis for the modeled network topology. Specific to the present project,
modeling based on a scorched node topology would mean that the location and number of
BTSs, Node Bs, BSCs, radio network controllers (RNCs), MSCs, and other equipment aregiven. Alternatively, scorched earth is an approach where the location and number of network
nodes are determined based on an optimal network design, taking into account current and
future demand levels.
Consistent with the MOCs prior cost modeling efforts, we determined that the model shouldhave the number of nodes that best reflects the actual networks in Israel. This modified
scorched node approach reflects international best practices and implies that the technology at
and in between existing switching nodes is optimized to meet the demands of a forward-looking efficient operator.9 Furthermore, such an approach takes into account:
The efficient choices of a hypothetical new entrant, which are constrained by the currentand future availability and commercial conditions associated with site procurement.
Therefore, it relies upon statistics about the design of the actual operators networks aspredictors of the network design constraints faced.
Network design is complicated as it involves a very large number of factors and design
parameters, not all of which are known, and this can make optimal network designuncertain.
Networks develop over time in response to changes in forecasted demand in addition totechnological evolution and uncertaintynot the theoretical limits of efficiency.
9 European Union Independent Regulators Group, Principles of implementation and best practice
regarding FL-LRIC cost modeling, as decided by the Independent Regulators Group, November 24, 2000, p. 3.
8/3/2019 NERA MAY 2010
17/71
.!
NERA Economic Consulting 9
2.5.3 Definition of forward looking
The forward-looking, efficient market outcome philosophy, which informs the LRIC approach,
requires that the asset values in the model be valued using the costs of an efficient entrant,rather than the historic costs incurred by the incumbent mobile network operators (MNOs). In
practice, this means that assets are valued based on the costs of replacing them with modernequivalent assets (MEAs). An MEA is the lowest-cost asset that serves the same function as the
asset being valued. It incorporates the latest available technology and is the asset that a new
entrant might be expected to employ.10 However, the model retains the flexibility to varynetwork capacities and traffic volumes for a set of different scenarios based on the input data
from Cellcom, Partner, and Pelephone.
In compliance with the MOCs stated goal of forecasting for a period of approximately five
years starting in 2009, the NERA model forecasts MEA prices and technological development
from 20092014, based on GSM, CDMA, and WCDMA technologies.11 Specifically, as shownin Table 2.1 below, the model produces network charges for the following operators and
technologies.
Table 2.1Modeled Operators and Technologies
Operator 2G Network 3G Network
Cellcom GSM 1800 WCDMA Release 7 dual band
Partner GSM dual band 900/1800 WCDMA Release 7 2100
Pelephone CDMA 850 WCDMA Release 7 dual bandSource: NERA
As Table 2.1 shows, each existing operator uses a different 2G technology and is modeled to
reflect this. Furthermore, based on the Israeli operators submissions and consistent with a
forward-looking modeling approach, NERA has modeled all three operators as havingWCDMA Release 7 3G networks. Specifically, in the case of Partner, this uses its 2100
frequency allocation. For Cellcom and Pelephone dual band assumptions are used.
2.6 Wholesale Access
The model produces network element costs which are aggregated to derive service costs.
Although the model has been designed with mobile termination in mind, with relatively minor
modifications this process can be extended to obtain the cost of wholesale access. Wholesaleaccess can take the form of either MVNO access or MNO access (i.e., national roaming).
10 European Union Independent Regulators Group, Principles of Implementation and Best Practice
Regarding FL-LRIC Cost Modelling, 24 November 2000, p. 611 SeeState of Israel, Ministry of Communications, Project Description: Consulting Services Regarding
Charges for Mobile Network Elements, May 2009, and the MOCs responses to clarification questions received
from bidding consulting firms.
8/3/2019 NERA MAY 2010
18/71
.!
NERA Economic Consulting 10
2.7 Study Process
In order to collect the necessary data for the NERA BU-LRIC model, NERA prepared a
comprehensive data request form in Microsoft Excel, along with a document describing the
necessary data for the model. This data request was submitted to Cellcom, Partner, Pelephone,and MIRS. Furthermore, following the submission of the data request, NERA had meetings
with all incumbent mobile operators in Jerusalem to clarify any questions they might have. We
also provided contact information to which questions regarding the data forms could be sent.
Partial data was submitted by the operators in several stages. NERA combined all data received
and examined the data for accuracy and consistency. We also had several rounds ofclarifications with some of the operators to ensure the proper understanding and use of the data.
The resulting data was then used as source data for the NERA BU-LRIC model.
However, despite NERAs comprehensive data request and efforts, considerable data were notprovided by the operators. Hence, where data were not provided by the operators, or if NERAconsidered the data received to be inconsistent with other data, we used data from the MOC.
Where no data were available from the MOC, we drew upon our previous experience of mobile
costing work and employed commercially and publicly available data from other jurisdictions.
8/3/2019 NERA MAY 2010
19/71
.!
NERA Economic Consulting 11
3 The NERA BU-LRIC Model
Most generally, the NERA BU-LRIC model involves the following tasks for each relevanttechnology.
1. Service demand assessment (What demand will the network serve?): This is the startingpoint of the NERA BU-LRIC model. It forecasts the number of subscribers and the traffic
volume for each type of service during the study period (20092014). Company-specificdemand levels are then calculated based on current and forecasted market shares.
2. Network dimensioning (What network components, size, and quantity are required?): Basedon the forecast demand, the model calculates the current and future capacity of the networkfor each technology based on the coverage requirement, traffic distribution, quality of
service, and so on. Next, the model calculates the physical quantities of components
required given the capacity of each network element (e.g., the number of subscribers ahome location register can serve).
3. Network asset valuation (What is the current and future cost of the network components?):Having established the required quantities of network equipment, the associated capital
expenditures are calculated by applying MEA unit cost and price trends. The model also
adds an allowance for indirect network assets such as buildings, vehicles, computers, andoffice equipment.
4. Capital charge determination (What is the annual capital charge for the networkinvestment?): Network capital expenditures are incurred to provide mobile services. Toderive the annual costs of network components and hence services, it is necessary to
annualize the cost of network assets, using an appropriate depreciation method and the
required return on capital.
5. Calculation of total costs (What is the total cost of operating each network element?): Themodel calculates operating expenditures for each type of equipment to capture the cost ofnetwork maintenance and operation and support. It then marks up the operatingexpenditures to allow for indirect network operating costs. The total (direct and indirect)
operational expenditures are added to the capital charge to arrive at the total LRIC cost.
6. Unit LRIC costs (What are the long-run average incremental costs allocated to specificservices and network elements?): In this final step, the model computes the unit LRIC cost
of each service. This involves the use of routing factors. For example, the average
requirements of an incoming call can be defined in terms of the number of base stations thecall passes through, the number and length of the transmission links, the number of
switching stages, and so on. These routing factors are then multiplied by the volume of each
service. The volume-weighted routing factors, in turn, are used to allocate the cost of eachof these network elements, including both capital and operating costs, in order to derive the
total network cost per minute for an incoming call.
The general NERA mobile costing approach is summarized in Figure 3.1.
8/3/2019 NERA MAY 2010
20/71
.!
NERA Economic Consulting 12
Figure 3.1Overview of NERAs BU-LRIC Model
Service demand assessment
Network dimensioning
Network asset valuation
Unit LRIC costs
Calculation of total LRIC
Subscriber forecasts
Traffic forecasts
Market shares
Busy hour traffic dimensioning
Geographic traffic distribution
Radio network
Quality of service
Engineering design rules (including network utilization)
Asset unit costs
Indirect non-network assets
Modern equivalent asset price trends
Direct operational expenses of network elements
Indirect operational expenses of network elements
Network common costs
Common costsMark up method
Routing factors
Service volumes
Capital charge determination Depreciation methodAsset lives
Return on capital employed (ROCE)
Model Conceptual Steps Inputs and Assumptions
Source: NERA
It is important to note that:
Where possible, we based the source data for our model on the data received from theIsraeli operators as part of the data gathering process and several subsequent clarification
rounds. We also incorporated relevant information and data received in meetings betweenNERA representatives and the representatives and consultants of all Israeli mobile
operators. However, as previously indicated, despite NERAs comprehensive data request,
considerable data were not provided by the operators. Hence, where data were not providedby the operators, or if NERA considered the data received to be inconsistent with other
data, we used data from the MOC. Where no data were available from the MOC, we drew
upon our previous experience of mobile costing work and employed commercially andpublicly available data from other jurisdictions. Specific examples of when we were not
able to use operator data, or chose not to, are provided in the relevant sections below.
8/3/2019 NERA MAY 2010
21/71
.!
NERA Economic Consulting 13
The costs of an individual service cannot be modeled in isolation of other services. This isbecause a large number of network components are used by more than one service (for
example a BTS is used to carry voice, SMS, and GPRS data services). Therefore, we have
included a comprehensive set of services in the model.
The model provides a simulation of the mobile networks in Israel, but there is a limit to thenumber of network elements that can be reasonably modeled. Thus, we have deliberately
limited the model to the most relevant and distinguishable network components rather than
attempt to model every single component of a mobile network.
Where possible, we compared the quantities of network elements contained in the modelwith those actually in the operators networks.
3.1 Service Demand Assessment
The first stage in deriving LRIC costs is to estimate the amount of capacity required to handle
the subscribers and market traffic volumes (i.e., the demand) in Israel during the study period.
Specifically, as shown in the 1 Inputs sheet of the NERA BU-LRIC model, we forecast thefollowing for the period 20092014:
Total mobile subscribers in Israel
Operator specific mobile subscribers
Operator market shares
Annual traffic splits
Market traffic volumes
In forecasting the demand levels, NERA relied on data received from the operators, the MOC,
and publicly and commercially available databases. The specific methodologies used to
forecast each variable are described in Appendix B to this report.
The service demand assessment commences with forecasting the total number of mobile
subscribers from 20092014, based on the population and mobile penetration rates in Israel.This is summarized in Table 3.1 below.
Table 3.1
Inputs Used to Forecast Subscribers in Israel20092014
2009 2010 2011 2012 2013 2014
Population 7,505,503 7,639,351 7,775,586 7,914,251 8,055,388 8,199,043
Penetration 124% 126% 128% 130% 132% 134%
Percent 3G 29.5% 32.8% 36.6% 40.9% 45.9% 51.5%
Source: NERA
8/3/2019 NERA MAY 2010
22/71
.!
NERA Economic Consulting 14
Next, based on data received from the operators, we forecasted market shares which allowed usto derive the operators respective traffic volumes12. Specifically, we estimated the 2009 market
shares based on the operators declarations. Market shares for 2010 were forecasted using
operator-specific average-growth rates in market share between 2007 and 2009. We assumedthat the MIRS market shares remain unchanged during the study period, and by 2016 a new 3G
entrant will obtain 10 percent of the residential market segment and 5 percent of the business
market segment. We split up market shares into voice, SMS, MMS and data market shares. Weunderstand that these assumptions are in line with general expectations. We further assumed
that the market share of the 3G entrant would come at the expense of all the other major
operators with each losing an equal percentage of its market share to the entrant each year. The
forecasted market shares are illustrated in Figure 3.2 and shown in Tables 3.2 to 3.4 below.
Figure 3.2Market Shares of Operators in Israel
19992014(E)
0%
10%
20%
30%
40%
50%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009E 2010E 2011E 2012E 2013E 2014E
MarketShare
Cellcom Pelephone Partner MIRS 3G Entrant
Source: NERA
12 Specifically, we created separate market shares for voice, SMS and also MMS and data. Due to datalimitations, we assumed the MMS and data markets followed Israel-wide market share trends, shown in Figure
3.2.
8/3/2019 NERA MAY 2010
23/71
.!
NERA Economic Consulting 15
Table 3.2Voice Market Shares of Operators in Israel
2009-2014(E)
2009E 2010E 2011E 2012E 2013E 2014E
Cellcom 33.3% 33.4% 32.8% 32.2% 31.7% 31.1%
Partner 33.4% 33.4% 32.8% 32.3% 31.7% 31.1%
Pelephone 28.2% 28.1% 27.6% 27.0% 26.4% 25.8%
MIRS 5.1% 5.1% 5.1% 5.1% 5.1% 5.1%
3G Entrant 0.0% 0.0% 1.7% 3.5% 5.2% 6.9%
Source: NERA
Table 3.3
SMS Market Shares of Operators in Israel2009-2014(E)
2009E 2010E 2011E 2012E 2013E 2014E
Cellcom 31.0% 31.1% 30.5% 29.9% 29.3% 28.8%
Partner 42.8% 42.9% 42.3% 41.7% 41.2% 40.6%
Pelephone 21.6% 21.5% 20.9% 20.4% 19.8% 19.2%
MIRS 4.6% 4.6% 4.6% 4.6% 4.6% 4.6%
3G Entrant 0.0% 0.0% 1.7% 3.5% 5.2% 6.9%
Source: NERA
Table 3.4MMS and Data Market Shares of Operators in Israel
2009-2014(E)
2009E 2010E 2011E 2012E 2013E 2014E
Cellcom 34.7% 34.8% 34.2% 33.6% 33.1% 32.5%
Partner 32.1% 32.2% 31.6% 31.0% 30.4% 29.9%
Pelephone 28.1% 28.1% 27.5% 26.9% 26.3% 25.8%
MIRS 5.0% 5.0% 5.0%` 5.0% 5.0% 5.0%
3G Entrant 0.0% 0.0% 1.7% 3.5% 5.2% 6.9%Source: NERA
The operator volumes were then further disaggregated into 2G and 3G traffic volumes, usingthe forecasts in Table 3.5 below. Starting with the traffic splits as declared by the operators, we
assumed that the proportion of traffic over each operators 2G network would decrease in the
same proportion as the percentage of 2G subscribers in the total market. This percentage, inturn, is calculated based on the number of 3G subscribers in the total market.
8/3/2019 NERA MAY 2010
24/71
.!
NERA Economic Consulting 16
Table 3.5Forecast 2G/3G Split in Israel
20092014
2009 2010 2011 2012 2013 2014
2G 75% 71% 67% 63% 58% 52%Cellcom
3G 25% 29% 33% 37% 42% 48%
2G 51% 49% 46% 43% 39% 35%Partner
3G 49% 51% 54% 51% 61% 65%
2G 60% 57% 54% 50% 46% 41%Pelephone
3G 40% 43% 46% 50% 54% 59%
Source: NERA analysis based on operator submissions
The forecasted percentage of 3G subscribers is the result of an econometric analysis based onmigration patterns and other attributes of 47 countries for which data were available around the
world. This model predicts the following evolution of 3G in Israel.
Figure 3.3Predicted Share of 3G in Israel
20092014
0%
10%
20%
30%
40%
50%
60%
70%
80%
2008 2009 2010 2011 2012 2013 2014
Percentage 3G
Percentage 2G
Source: NERA analysis
Having forecast the total number of subscribers, the market shares of different operators, and
the traffic split between 2G and 3G, we forecasted the market traffic volumes. This involved
8/3/2019 NERA MAY 2010
25/71
.!
NERA Economic Consulting 17
estimating total call minutes and total successful calls for each of the following wirelessservices:
Mobile to mobile on-net
Mobile to mobile off-net
Mobile to fixed
Mobile to international
Fixed to mobile (termination)
Other to mobile (termination)
Mobile to voicemail
Fixed to voicemail
International roaming
We also forecasted the volumes of SMS, MMS, GPRS, and WCDMA traffic. To convert SMS
traffic into a voice-minute equivalent, we first determined the average size of an SMS as 154
bytes. This consists of 70 bytes for an average SMS message and 84 bytes for the networkoverhead for an SMS. These figures were provided by the Israeli MNOs. SMSs travel on the
networks control channels, which have a modeled capacity of 9,600 bits per seconds. There
are eight bits in a byte, making a standard SMS equal to 1,232 bits. As only two of the 25
frames accommodate SMS traffic, there are 768 bits of capacity per second available forSMSs.13 Hence, a standard SMS has a voice-equivalent of 1.60 seconds.14 Stated differently,
one voice minute is equivalent to 37 SMSs.15 This conversion is illustrated in the figure below.
13 (9,600 bps/25 frames)*2frames=768 bps14 1,232 bits/768 bps=1.60 seconds15 60 second/1.60 seconds=37.5 SMS per minute
8/3/2019 NERA MAY 2010
26/71
.!
NERA Economic Consulting 18
Figure 3.4SMS Traffic Conversion Process
Average size of an SMS overhead[source Israel MNO, 84 Bytes]
Average size of an SMS overhead[source Israel MNO, 84 Bytes]
Average size of an SMS data traffic
[source Israel MNO, 70 Bytes]
Average size of an SMS data traffic
[source Israel MNO, 70 Bytes]
Control channel sizes
[source NERA, 768 bps, 2 of the 25 frames in
the 9,600 bps channel carry SMS]
Control channel sizes
[source NERA, 768 bps, 2 of the 25 frames in
the 9,600 bps channel carry SMS]
Bits in a Byte [8]Bits in a Byte [8]
Seconds in a minute [60]Seconds in a minute [60]
Total size of an SMS [154 bytes]Total size of an SMS [154 bytes]
SMS equivalent to 1 voice minute [37]SMS equivalent to 1 voice minute [37]
Average size of an SMS overhead[source Israel MNO, 84 Bytes]
Average size of an SMS overhead[source Israel MNO, 84 Bytes]
Average size of an SMS data traffic
[source Israel MNO, 70 Bytes]
Average size of an SMS data traffic
[source Israel MNO, 70 Bytes]
Control channel sizes
[source NERA, 768 bps, 2 of the 25 frames in
the 9,600 bps channel carry SMS]
Control channel sizes
[source NERA, 768 bps, 2 of the 25 frames in
the 9,600 bps channel carry SMS]
Bits in a Byte [8]Bits in a Byte [8]
Seconds in a minute [60]Seconds in a minute [60]
Total size of an SMS [154 bytes]Total size of an SMS [154 bytes]
SMS equivalent to 1 voice minute [37]SMS equivalent to 1 voice minute [37]
Source: NERA
Similarly, to put MMS traffic on a common basis with other traffic, we first determined the
average size of an MMS as 70,000 bytes, using data provided by the operators in Israel. Thisconsists of 10,000 bytes for MMS overhead and 60,000 bytes for the actual MMS. An MMS
travels on the networks data channels, merging with other traffic, which is measured in
megabytes. Converting total MMSs into megabytes results in 0.070 MBs (megabytes). Hence, astandard MMS is equivalent to 0.070 MBs of data traffic. This process is illustrated in the
figure below.
8/3/2019 NERA MAY 2010
27/71
.!
NERA Economic Consulting 19
Figure 3.5MMS Traffic Conversion Process
Average size of an MMS message[source Israel MNO, 60,000 Bytes]Average size of an MMS message[source Israel MNO, 60,000 Bytes]
Average size of an MMS overhead
[source NERA, 10,000 Bytes]
Average size of an MMS overhead
[source NERA, 10,000 Bytes]
Total size of an MMS [70,000 bytes]Total size of an MMS [70,000 bytes]
MMS demand in Megabytes (MB) [0.070]MMS demand in Megabytes (MB) [0.070]
Average size of an MMS message[source Israel MNO, 60,000 Bytes]Average size of an MMS message[source Israel MNO, 60,000 Bytes]
Average size of an MMS overhead
[source NERA, 10,000 Bytes]
Average size of an MMS overhead
[source NERA, 10,000 Bytes]
Total size of an MMS [70,000 bytes]Total size of an MMS [70,000 bytes]
MMS demand in Megabytes (MB) [0.070]MMS demand in Megabytes (MB) [0.070]
Source: NERA
3.1.1 Voice and data demand equivalence
Both voice and data use resources in the radio network, although once away from the radio
network voice traffic and data traffic are handled and routed separately through the core
network. Voice demands are expressed in minutes and calls, and data demands are expressed inmegabytes. In the model, we align these two demands by applying a factor to the data demand
to represent the equivalent intensity in voice minutes. The factor is derived by considering thecapacity limits of a cell in terms of call minutes and in terms of megabytes of data. The
relationship between the amount of call minutes that exhaust a cells capacity and the amount
of megabytes that exhaust a cells capacity is the equivalence factor. The factor differs for 2G
GSM, 2G CDMA 1x, and WCDMA release 7 because each of these has different voice-minutecapacities and different cell data-rate capacities. We outline the specific conversion process
below.
First, we establish the cell-voice capacity in Erlangs and then convert that to annual minutes.
This conversion entails converting Erlangs to call minutes in the busy hour and then convertingthe busy hour call minutes to an annual equivalent. This gives us the annual minutes that a cell
can support. Second, we establish the cell data-rate capacity after allowing for IP overheads and
adjusting for the total traffic including both downlink and uplink; we then convert that to thedata transmitted in a year. An example using a WCDMA Release 7 technology is shown in
Figures 3.6 and 3.7 below.
8/3/2019 NERA MAY 2010
28/71
.!
NERA Economic Consulting 20
Figure 3.6WDCMA Busy Hour Data Capacity
Minimum channel rate in the busy hour[source MNOs, 4.5 Mbit/s]
Minimum channel rate in the busy hour[source MNOs, 4.5 Mbit/s]
Bits in a Byte [8]Bits in a Byte [8]
Seconds in an hour [3600]Seconds in an hour [3600]
Total Mbytes in the busy hour [2,260 Mbytes]Total Mbytes in the busy hour [2,260 Mbytes]
Downlink proportion [80%]Downlink proportion [80%]
IP overhead [12%]IP overhead [12%]
Source: NERA
The downlink (cell to mobile) data rate (achievable in the Israeli context using WCDMA
Release 7) is approximately 4.5 Mbps in busy traffic conditions. In a busy hour, that isequivalent to 16,200 Mbps (in the busy hour). As there are eight bits in a byte, the hourly datarate is 2,025 MBs in the busy hour. Again, assuming that annual traffic is 10 busy hours on 250
busy days, this equates to annual downlink traffic of 5,062,500 MBs. We assume a downlink
proportion of stated demand to be 80 percent (this is another input in the model that can bealtered) and an IP overhead of 12 percent (also changeable in the assumptions). Adjusting for
these two parameters, the annual demand that would exhaust a cell capacity is 5,650,112 MBs.
8/3/2019 NERA MAY 2010
29/71
.!
NERA Economic Consulting 21
Figure 3.7WDCMA Busy Hour Voice and Data Equivalence
WCDMA r7 cell voice capacity[source MNOs, 69 channels]
WCDMA r7 cell voice capacity[source MNOs, 69 channels]
Erlang capacity 54.64
[at 1% GoS]
Erlang capacity 54.64
[at 1% GoS]
Minutes in an hour [60]Minutes in an hour [60]
Total minutes in the busy hour [3,278 minutes]Total minutes in the busy hour [3,278 minutes]
Megabytes of data in busy hour
[2,260]
Megabytes of data in busy hour
[2,260]Voice / data equivalence [1.45]
[in WCDMA r7, Israel morphology]
Voice / data equivalence [1.45]
[in WCDMA r7, Israel morphology]
Source: NERA
The voice channel capacity for WCDMA release 7 is assumed to be 69. (The data were
provided by Israeli operators and is alterable in the model so that other assumptions can beemployed if desired.) At a Grade of Service of 1 percent, 69 channels will provide a capacity of
54.639 Erlangs of carried traffic. This is equivalent to 54.639 x 60, 3,278 minutes of traffic inthe busy hour. Assuming that a year is equivalent to 10 busy hours on 250 busy days, thisequates to 8,195,864 minutes (when using full precision values for the Erlang traffic capacity).
The ratio between 8,195,864 voice minutes, and 5,650,112 megabytes, or 1.45, is the voice anddata equivalence factor. That is, 1.45 annual voice minutes exhaust the same capacity of a cell
as 1 MB of data annually.
3.1.2 Using demand figures
Based on total voice-equivalent traffic estimates, we calculated the required network capacity
for each operator and technology. For Cellcom, we dimensioned an 1800 MHz GSM networkand a dual band 850/2100 MHz WCDMA Release 7 network. For Partner, we dimensioned a900/1800 MHz dual band GSM network and a 2100 MHz WCDMA network. For Pelephone,
we dimensioned an 850 MHz CDMA network and a dual band 850/2100 MHz WCDMAnetwork.
The network dimensioning process involved the following generic steps:
1. Include an allowance for unsuccessful calls (call attempts) and ringing time (call minutes)in the voice-equivalent network demand determined above.
8/3/2019 NERA MAY 2010
30/71
.!
NERA Economic Consulting 22
2. Multiply total voice minutes and total call attempts by equipment usage factors to obtainequipment minutes (or attempts). For example, a terminating mobile call makes use of one
BTS, whereas an on-net call makes use of two BTSs.
3. Use the ratio of minutes in the peak hour to minutes over the year to estimate busy hourErlangs (BHE) and busy hour call attempts (BHCA).16
4. Use an Erlang B table together with the assumed blocking probability to estimate thenumber of channels required to handle the busy hour traffic, taking into account modularity.
In the data request, we asked for traffic-parameter information from the operators to use in thisprocess. The responses are summarized in Table 3.6 below.
Table 3.6Traffic Parameters
Parameter (units) Cellcom Partner Pelephone
Average conversation time (minutes) 1.81 1.40 1.19
Average non-conversation holding time (minutes) 0.30 0.30 0.23
Completed calls (% of all call attempts) 99.0% 56.0% 71.4%
Source: Operator submissions
The operator submissions above appeared to be inconsistent. For instance, the average
conversation time (minutes) varied significantly across operators and, in some cases, were
inconsistent with declarations of call minutes. Consequently, we used a market value instead
that was based on actual conversation minutes and call attempts in the Israeli market. Similarly,average-nonconversation holding times seemed high with Cellcom and Partner each reporting0.3 minutes (18 seconds). Hence, we used the figure received from Pelephone instead, which is
0.23 minutes (13.8 seconds) and more in line with our experience in other countries.
Finally, completed calls (as a percentage of all call attempts) varied substantially between thedifferent operators. With 99 percent completed calls, Cellcoms submission appears
anomalously high relative to international benchmarks, whereas Partners 56 percent seems
low. Hence, we used Pelephones figure of 71.4 percent instead as it was consistent withexpectations and what we have found in other countries.
3.2 Network DimensioningBased on demand and the engineering principles and algorithms that determine the required
network capacities, the NERA model determines the number of physical units of all network
elements. This network dimensioning module determines the volume of network elementsrequired to support the given level of demand using the technology chosen. Specifically,
16 There are different theories of congestion in telephone networks, which deal with the ability of
facilities to handle loads that may be imposed on them. The Erlang B table uses a formula that calculates the
number of facilities required when a maximum load is present based on the assumptions that an infinite number ofsources exist, calls arrive randomly and are served in the order of arrival, blocked calls are lost, and holding times
are distributed exponentially.
8/3/2019 NERA MAY 2010
31/71
.!
NERA Economic Consulting 23
network elements are grouped into three categories (radio network, core network, andtransmission network). For each technology (i.e., GSM Dual 800/900-1800, GSM 1800,
CDMA 850, and WCDMA Release 7), a network covering Israels territory was designed to
handle the level of demand faced by the operator. An illustrative example of such a topologydesign (for GSM) and its corresponding network elements is shown in Figure 3.8.
Figure 3.8Illustrative Simplified Network Topology
GSM
TRXGSM Transceiver
BTSGSM Macro Base Station
MICROGSM Micro Base Station
PICOGSM Pico Base Station
BScBase Station Controller
MSCMobile Switching Centre
TSTandem Switch
SGSNinterface between the GPRS network
GGSNinterface between the GPRS
network and external IP networks
POIPoint of interconnect
MSC
BTS
POI
Prepay platform
Short Message Service Platform
Mobile Multimedia database
Voicemail Server
Interconnect billing system
Network management system
Home location register
TRX
TRX TRX
TRX
TRX TRX
BTS
TRX
TRX TRX TRX
TRX TRXTRX
TRX TRX
SGSN
MSC
TS
PRP
SMS
MMS
VMS
BIL
NMS
HLR
GPR GPRS Software
TRX
TRX
TRX
PICO POI
TSBSC
MICRO
BSC
GGSN
TRXGSM Transceiver
BTSGSM Macro Base Station
MICROGSM Micro Base Station
PICOGSM Pico Base Station
BScBase Station Controller
MSCMobile Switching Centre
TSTandem Switch
SGSNinterface between the GPRS network
GGSNinterface between the GPRS
network and external IP networks
POIPoint of interconnect
MSC
BTS
POI
Prepay platform
Short Message Service Platform
Mobile Multimedia database
Voicemail Server
Interconnect billing system
Network management system
Home location register
TRX
TRX TRX
TRX
TRX TRX
BTS
TRX
TRX TRX TRX
TRX TRXTRX
TRX TRX
SGSN
MSC
TS
PRP
SMS
MMS
VMS
BIL
NMS
HLR
GPR
PRP
SMS
MMS
VMS
BIL
NMS
HLR
GPR GPRS Software
TRX
TRX
TRX
PICO POI
TSBSC
MICRO
BSC
GGSN
Source: NERA
The network topology for a CMDA network is similar to the GSM architecture shown above,
as the same network components are used.
The network architecture of a WCDMA network is also similar to GSM. However, although
the network components fulfill similar, albeit at times enhanced, functions, they have differentnames. Specifically, the WCDMA equivalent of a BTS is a Node B and a BSC is called an
RNC (a radio network controller). An RNC performs an enhanced role relative to that of aBSC. In addition to controlling the radio network, the RNC also controls handovers to other
Node Bs when a mobile subscriber changes locations. This, in turn, requires transmission links
between RNCs. There are several releases of WCDMA, starting with Release 98 in 1998.Recent releases of WCDMA technology include High Speed Download Packet Access
(HSDPA) in Release 5, High Speed Uplink Packet Access (HSUPA) in Release 6, and Voice
over Internet Protocol (VoIP) and High Speed Packet Access (HSPA+) in Release 7. These
8/3/2019 NERA MAY 2010
32/71
.!
NERA Economic Consulting 24
later releases require upgrades to the RNCs and Node Bs. These upgrade costs are reflected inthe cost model.
Although Releases 8, 9, and 10 are at various stages of the development process, the NERABU-LRIC model dimensions WCDMA networks for Release 7. We find this to be the
appropriate technology and architecture modeling choice for Israel for several reasons. First, intheir submissions, the Israeli MNOs state that WCDMA Release 7 is scheduled to be deployed
in 2010. Hence, it represents a forward-looking technology, consistent with the models
objective. Second, given the continued uncertainty in the industry over Release 8 and later
releases, it appears premature (and consequently controversial) to derive cost and performanceestimates for these releases. Finally, the Israeli MNOs also have indicated a preference to
deploy Release 10 (LTE) when it becomes available.17
3.2.1 Model geotypes
In order to tailor the modeled networks to the different levels of demand in varied geographic
areas, the NERA BU-LRIC model categorizes the demand according to traffic densities indifferent areas (known as geotypes). The geotype information provided by the operators is
shown in Table 3.7 below.
Table 3.7Geotype Characteristics
Cellcom Partner Pelephone
Geotype Area
(sq.km)
Population Area
(sq.km)
Population Area
(sq.km)
Population
Dense
Urban- - - - 158 -
Urban 1,857 5,990,580 1,824 5,972,000 4,698 -
Suburban 3,153 836,510 24,619 1,493,000 2,844 -
Rural 16,650 18,250 - - 19,034 -
Deserted - - 1,410 - 885 -
Total 21,659 6,845,340 27,853 7,465,000 27,619 -
Source: Operator submissions Note: - denotes no data submitted.
As can be seen, the three operators seem to have relied on different definitions of geotypes.Cellcom divided its serving area into three categories (urban, suburban, and rural), Partner used
three differently defined categories of geotype (urban, suburban, deserted), and Pelephone used
five geotypes. Cellcoms data were problematic in that the area figures appeared to exclude the
West Bank, although other data from Cellcom did not. Similarly, Partners data were difficultto incorporate as no information was provided about rural areas. Hence, we opted to rely on
17 LTE is a mobile communication system that is an evolution of the GSM and UMTS systems. It is still a
3G technology. LTE Advanced is a 4G technology.
8/3/2019 NERA MAY 2010
33/71
.!
NERA Economic Consulting 25
Pelephones data, combining dense urban and urban into one geotype. The resultinggeotype data, which are used in the model, are shown in Table 3.8 below. We note that,
although deserted areas are taken into account, the model does not dimension for these areas as
there is no traffic in this geotype.
Table 3.8Geotype Data Used in Model
Area (km) % of Total Area % of Annual Traffic
Urban 4,856 17.6% 81.7%
Suburban 2,844 10.3% 6.8%
Rural 19,034 68.9% 11.5%
Deserted 885 3.2% 0.0%
Total 27,619 100.0% 100.0%
Source: NERA analysis based on operator submissions
The size of the network in each geotype is determined by one of two drivers. Either the
dimensioning is based on the coverage requirement (i.e., the number of BTSs required toprovide geographic coverage), or the size is determined by the network traffic in the area (i.e.,
the network must provide a given level of traffic capacity for a given level of utilization).
3.2.2 Dimensioning assumptions
Network dimensioning starts with estimating the radio network required to handle demand. We
assume three different cell types: macrocells, microcells, and picocells. Although the area that
can be covered by each cell type depends on the immediate environment, macrocells have thelargest cell radius, followed by microcells, and then picocells. For macrocells, the base station
antennas are installed on a mast (greenfield) or on top of a tall building. Microcells require
lower antenna heights and are typically used in urban settings. Finally, a picocell is designed to
serve a very small area, such as part of a building, a street corner, or an airplane cabin. Theycan be used to extend coverage to indoor areas that are inaccessible to outdoor signals or to add
network capacity in areas with very dense phone usage.
The objective of the model is to design a radio network configuration that meets the required
level of demand. To do this, we incorporated the following assumptions in the model:
A network is rolled out to provide geographical coverage (i.e., there is only a minimumlevel of traffic).
Each coverage base station contains one sector, and there is a minimum of one transceiverper sector.
The minimum transceiver configuration of a macro base station is one transceiver persector.
8/3/2019 NERA MAY 2010
34/71
.!
NERA Economic Consulting 26
To accommodate traffic, MNOs usually have to split sites into several smaller areas tohandle the density of traffic. The model tracks that process by employing the cell radii that
the MNOs in Israel have found to be necessary, thus providing enough cell sites to handle
the traffic. This approach has the additional benefit that the particular mix of terrain andbuildings in Israel is reflected in the cell radii actually employed by the operators.
Transceivers are added to each base station in response to traffic demand until each basestation is fully configured.
In the dual band networks, such as GSM 900-1800 or WCDMA 850-2100, transceivers foreach band are colocated at the same base stations and additional GSM 1800 transceivers (or
WCDMA 850 and 2100 transceivers) and equipment are added to provide additional trafficcapacity.
Once each base station is fully configured with both GSM 900 and GSM 1800 transceivers
(or WCDMA 850 and 2100 transceivers), additional base stations are added to provide
additional traffic capacity.
The upper limit on the number of transceivers per base station is determined by either:
The physical limit of the number of transceivers, which is a maximum of sixtransceivers per sector.
The number of transceivers per sector that the spectrum will allowthe modelderives this from the spectrum allocations and, for bands that can be used for both
2G and 3G (such as 850), after allowing for any allocation to either 2G or 3G.
2G and 3G networks for each existing MNO are modeled as if it was entering themarket now at its current scale and scope of operation with either a single 3G network
or both a 2G and 3G network.
3.2.3 Minimum coverage network
Based on these assumptions, the model derives a minimum coverage network. Specifically, itdetermines the number of base stations required to provide coverage to the area and the
minimum amount of network equipment required to enable a voice call to be conveyed between
any two points in the coverage area. It is important to note that the costs of the minimumcoverage network are a network common cost (i.e., the minimum coverage network is required
whichever services are provided. Its annualized cost is calculated and treated as a markup on
the LRIC costs of the traffic network. A summary of the volumes of network elements requiredfor the minimum coverage network for Cellcoms GSM 1800 network is shown in Table 3.9.
8/3/2019 NERA MAY 2010
35/71
.!
NERA Economic Consulting 27
Table 3.9Minimum Coverage Network
Example of Cellcom GSM 1800
Network Element 2009 2010 2011 2012 2013 2014
BTS 553 553 553 553 553 553
BSC 2 2 2 2 2 2
MSC 1 1 1 1 1 1
HLR 1 1 1 1 1 1
TS 1 1 1 1 1 1
Source: NERA calculations
Note: These values depend on the scenario chosen.
We also include an interconnection billing system, a point of interconnection, and spectrum
license fees in the definition of the minimum coverage network, as these are required to providevoice services. The minimum network definition excludes the provision of prepaid, SMS, and
GPRS data services because they would not be provided as part of the minimum coverage
network requirement.
3.2.4 Traffic network
While the minimum coverage network determines the common cost component, the traffic
network determines the incremental costs.
3.2.4.1 Cell radii and base station types
The cell radii for the traffic network in the NERA BU-LRIC model are summarized in Table
3.10.
Table 3.10Cell Radii for Traffic Base Stations (km)
Geotype2G
CDMA850
2GGSM900
2GGSM1800
3GWCDMA
850
3GWCDMA
900
3GWCDMA
2100
Urban 1.5 1.5 1.5 1.5 1.5 1.3
Suburban 2.2 3.5 3.5 2.2 2.2 1.8
Rural 4.3 4.0 4.0 4.3 4.3 3.6
Highway/Railway 2.2 3.5 3.5 2.2 2.2 1.8
Tunnel etc 1.5 1.5 1.5 1.5 1.5 1.3
Source: NERA analysis
Unlike in the case of a minimum coverage network, the traffic network also includes microcells
and in-building base stations. However, these are deployed only in urban and suburban areas.
8/3/2019 NERA MAY 2010
36/71
.!
NERA Economic Consulting 28
We dimension the amount of traffic carried by these base stations using information in theoperators submissions. As Pelephone did not provide any data in response to this request, we
used for Pelephone the data provided by Partner as these data were most complete.
Table 3.11Traffic Carried Using Micro- and Picocells
Operator Technology Microcell Picocell
2G 3% n/aCellcom
3G 3% n/a
2G 2% 5%Partner
3G 0% 4%
2G - -Pelephone3G - -
Source: Operator submissions Note: - denotes no data submitted.
The proportion of different cell-site types and the way in which the equipment is mounteddiffers by operator. For Cellcom and Partner, the distribution of cell-site types relies on the
carriers respective data declarations. Pelephone did not submit any data. Consequently,
Pelephones cell-type mix is based on data received from the Ministry of EnvironmentalProtection, which lists all cell sites in Israel based on cell-site type.
The NERA model also allows for areas of special coverage, specifically along highways or
railways and in tunnels, underground, subway or metro systems. Data in response to NERAsrequest on this point were received from Cellcom and Partner. However, the Cellcom
Highway/Railway figures appeared anomalously high, so we have used those from Partnerinstead. Conversely, Partner provided no Tunnel/Subway/Underground/Metro data, so we
have relied on the Cellcom figures.
Table 3.12Special coverage
Type Coverage (km)
Highway/Railway 26
Tunnel/Subway/Underground/Metro 100
Source: NERA
3.2.4.2 Traffic capacity calculation
The traffic capacity required by each base station is determined using an Erlang B table. An
Erlang is a unit of telecommunications traffic measurement, which represents the continuoususe of one voice path (and thus the related traffic volume) for one hour. Erlang traffic
measurements or estimates can be used to work out how many circuits are required between
8/3/2019 NERA MAY 2010
37/71
.!
NERA Economic Consulting 29
different parts of a network or between multiple network locations. The traffic capacity (asshown in the Erlang B table) is a function of:
The number of transceivers deployed in each base station sector
The number of traffic channels per transceiver
The level of call blocking in the radio network
The model includes an assumed blocking probability of 1 percent (or P=.01) for 2G and 3G
networks. This is consistent with the data submission from Cellcom and general engineeringstandards. An example of this traffic dimensioning calculation is shown in Figure 3.9 below.
Figure 3.9Traffic Dimensioning Calculation
# transceivers per sector
e.g. 3 to provide maximumcapacity on the network
within the spectrum allocation
# transceivers per sector
e.g. 3 to provide maximumcapacity on the network
within the spectrum allocation
Number of channel
per sector
e.g. 207
Number of channelper sector
e.g. 207
# channels per transceiver
Set to 69 channelsper 3G transceiver
[typical value for WCDMA]
# channels per transceiver
Set to 69 channelsper 3G transceiver
[typical value for WCDMA]
Level of radio network blocking
e.g. 1%[% of calls that cannot be made
because a radio channelis not available]
Level of radio network blocking
e.g. 1%[% of calls that cannot be made
because a radio channelis not available]
Erlangs per sectore.g. each sector can carry
184.6 Erlangs of traffic
The Erlang table is used tocalculate the amount of
traffic that can be carried byeach sector
Erlangs per sectore.g. each sector can carry184.6 Erlangs of traffic
The Erlang table is used tocalculate the amount of
traffic that can be carried byeach sector
Erlang table look-up
# transceivers per sector
e.g. 3 to provide maximumcapacity on the network
within the spectrum allocation
# transceivers per sector
e.g. 3 to provide maximumcapacity on the network
within the spectrum allocation
Number of channel
per sector
e.g. 207
Number of channelper sector
e.g. 207
# channels per transceiver
Set to 69 channelsper 3G transceiver
[typical value for WCDMA]
# channels per transceiver
Set to 69 channelsper 3G transceiver
[typical value for WCDMA]
Level of radio network blocking
e.g. 1%[% of calls that cannot be made
because a radio channelis not available]
Level of radio network blocking
e.g. 1%[% of calls that cannot be made
because a radio channelis not available]
Erlangs per sectore.g. each sector can carry
184.6 Erlangs of traffic
The Erlang table is used tocalculate the amount of
traffic that can be carried byeach sector
Erlangs per sectore.g. each sector can carry184.6 Erlangs of traffic
The Erlang table is used tocalculate the amount of
traffic that can be carried byeach sector
Erlang table look-up
Source: NERA
Figure 3.9 illustrates how the NERA BU-LRIC model dimensions the traffic on one base
station sector. In this example, there are three transceivers per sector. Each WCDMA
transceiver on a sector can be configured for 69 voice channels. For three transceivers, there are207 (69 x 3) channels available to accommodate the traffic. Using a blocking probability of 1
percent, the Erlang table indicates that with 207 channels, 184.6 Erlangs of traffic can becarried by the sector. As illustrated in this example, the capacity (Erlangs per sector) also
depends on the number of transceivers per sector. This, in turn, is calculated using:
The level of traffic capacity required
The traffic capacity provided by the traffic network
The cap on the number of transceivers per base station resulting from either the amount ofavailable spectrum or the physical capacity per sector (which varies depending on the
technology employed)
8/3/2019 NERA MAY 2010
38/71
.!
NERA Economic Consulting 30
The Erlang look-up table that provides the amount of traffic that can be handled for a givenblocking probability and the number of available channels is based upon statistical engineering
calculations. An excerpt from it is shown in Table 3.13 below.
Table 3.13Excerpt of Erlang Lookup Table
Blocking ProbabilitiesChannels
0.10% 0.50% 1% 2% 3% 5% 10%
0 0.00 0.00 0.00 0.00 0.00 0.00 0.00
1 0.00 0.01 0.01 0.02 0.03 0.05 0.11
2 0.05 0.11 0.15 0.22 0.28 0.38 0.60
3 0.19 0.35 0.46 0.60 0.72 0.90 1.274 0.44 0.70 0.87 1.09 1.26 1.52 2.05
Source: ITU
3.2.4.3 Spectrum reuse patterns
Before the traffic network can be dimensioned, it is necessary to determine not only the cell
radius but also the frequency reuse pattern in the case of multifrequency networks, such as
GSM. In contrast to a network with a single transmitter, a cellular network has multipletransmitters and thus the ability to reuse the same frequency in a different area. This increases
the capacity in a cellular network because many subscribers can use a single frequency. In
order to minimize the level of interference from reusing the same frequency, a frequency reusepattern must be established. The frequency reuse pattern sets an upper limit on the number of
transceivers that a base station sector can operate for a given amount of spectrum. This, in turn,
influences the number of base stations in the network. It should be noted that WCDMAoperates as a single frequency network (SFN), where each base station operates at the same
frequency (or frequencies) as its neighbors; in this case, frequency reuse is 100 percent.
To determine the reuse pattern, the coverage area is depicted as a series of contiguous
hexagons. Each hexagon represents the serving area of a base station, which is split into threesectors. As shown in Figure 3.10, we assume a frequency reuse pattern of 12 for GSM, which
allows there to be at least one sector gap between cells that use the same frequency. This is
equivalent to a base station frequency reuse of four (assuming three sectors per base station).
8/3/2019 NERA MAY 2010
39/71
.!
NERA Economic Consulting 31
Figure 3.10Example of GSM Frequency Reuse Pattern
A1
A2
A3
A1
A2
A3
D1
D2
D3
D1
D2
D3B1
B2
B3
B1
B2
B3
D1
D2
D3
D1
D2
D3
A1
A2
A3
A1
A2
A3 B1
B2
B3
B1
B2
B3 C1
C2
C3
C1
C2
C3
A1
A2
A3
A1
A2
A3
C1
C2
C3
C1
C2
C3
C1
C2
C3
C1
C2
C3
A1
A2
A3
A1
A2
A3
C1
C2
C3
C1
C2
C3
B1
B2
B3
B1
B2
B3
D1
D2
D3
D1
D2
D3
C1
C2
C3
C1
C2
C3
B1
B2
B3
B1
B2
B3
A1
A2
A3
A1
A2
A3
A1
A2
A3
A1
A2
A3
Cell radiuse.g. 2km
Note:
Each hexagon represents a different basestation
There is a frequency re-use pattern of 12
The basestations have a frequency re-use factor of 4
The number of sectors per basestation is 3 Source: NERA
We further assume that each GSM TRX requires 200 kHz of paired spectrum and provideseight 25 kHz communication channels. For CDMA, we assume that a TRX requires 1.25 MHz
of paired spectrum and provides 35 25 kHz communication channels. For WCDMA, we
Top Related